Handling Apollo cache updates after creating and deleting objects, or
associating and dissociating objects, remains a
poorly solved problem.
update
and refetchQueries
props on Mutation
s couple different areas of
your app in a way you probably don't want, and they don't scale well as you add
more queries over objects you may create/delete.
Truly solving the problem will probably require changes to the apollo client and cache code.
Until that happens, this is probably your best bet!
- How it works
- Current limitations
- ES environment requirements
- Type metadata usage
- Handling Deletions
- Handling Creation
- Handling associations being broken
- Handling associations being created
- API
After you delete an object, you tell apollo-magic-refetch
what typename
and
id
was deleted, and it refetches all active queries that contain that object
anywhere within their current data!
Similarly, after you create an object, you tell it the typename
of the created
object and it refetches all active queries that contain an object of that type
in their selections. This is a bit less efficient than handling deletes, but
way easier than anything else at the time of writing.
Since only active queries can be refetched, data in the cache for inactive
queries will remain out-of-date. For that reason, I would recommend using the
cache-and-network
policy on all queries you're not planning to update
after
all pertinent mutations.
- Interfaces and union types are not supported yet. This means if they are anywhere in your results, this library may fail to refetch when it should.
- Lists of lists are not supported yet.
If you are building for legacy browsers with a bundler like Webpack, make sure to add a rule to transpile this package to ES5.
If you are not using a bundler that supports the module
property in
package.json
, make sure to install babel-runtime
.
apollo-magic-refetch
uses type metadata from GraphQL determine which queries
need to be refetched; the client must get this metadata from the server.
If your schema is large enough it may be a prohibitive amount of metadata.
refetch
operations will be delayed until this metadata is fetched.
To prefetch this metadata via a GraphQL introspection query, do:
import client from './wherever/you/create/your/apollo/client'
import refetch from 'apollo-magic-refetch'
// initiate the prefetch
refetch.fetchTypeMetadata(client)
If your server forbids client introspection queries, you will have to fetch the metadata by other means. For instance, you could execute the required introspection query on the server, and serve the result on a custom REST route:
import { execute } from 'graphql'
import schema from './path/to/your/graphql/schema'
import express from 'express'
import { typesQuery } from 'apollo-magic-refetch'
const app = express()
const typeMetadataPromise = execute(schema, typesQuery)
app.get('/graphql/refetchTypeMetadata', (req, res) => {
typeMetadataPromise.then((data) => res.json(data))
})
And then pass this data to refetch.setTypeMetadata
before you ever call
refetch()
:
import refetch from 'apollo-magic-refetch'
// accepts a promise that resolves to the graphql execution result.
refetch.setTypeMetadata(
fetch('/graphql/refetchTypeMetadata').then((res) => res.json())
)
Typically you call refetch
within the update
callback of your Mutation
that deletes objects. You just have to call refetch
with the __typename
that was deleted (in this case, Device
) and the id
of the deleted object.
This refetches any active queries that contain the deleted object in cached data.
For mutations that delete multiple things at once, you may pass an array or Set
of ids to refetch
, or make multiple calls to refetch
in your update
method.
import * as React from 'react'
import gql from 'graphql-tag'
import refetch from 'apollo-magic-refetch'
import {Mutation, ApolloConsumer} from 'react-apollo'
const mutation = gql`
mutation destroyDevice($deviceId: Int!) {
destroyDevice(deviceId: $deviceId)
}
`
const DestroyDeviceButton = ({deviceId}) => (
<ApolloConsumer>
{client => (
<Mutation
mutation={mutation}
update={() => refetch(client, 'Device', deviceId)}
/>
{destroyDevice => (
<button onClick={destroyDevice({variables: {deviceId}})}
)}
</Mutation>
)}
</ApolloConsumer>
)
Typically you call refetch
within the update
callback of your Mutation
that creates objects. You just have to call refetch
with the __typename
that was created.
Unlike deletions, you don't pass the id
of the created
object. Without a specific id
to search for, it simply refetches all active
queries that contain any object of the requested __typename
in their cached
data, in case the created object belongs in the new results. This is less
efficient than refetching queries containing a specific id
, but far easier
than manually inserting the created object into each relevant query.
In this example, the __typename
of the object being created is Device
.
import * as React from 'react'
import gql from 'graphql-tag'
import refetch from 'apollo-magic-refetch'
import {Mutation, ApolloConsumer} from 'react-apollo'
import CreateDeviceForm from './CreateDeviceForm'
const mutation = gql`
mutation createDevice($values: CreateDevice!) {
createDevice(values: $values) {
id
}
}
`
const CreateDeviceFormContainer = () => (
<ApolloConsumer>
{client => (
<Mutation
mutation={mutation}
update={() => refetch(client, 'Device')}
/>
{createDevice => (
<CreateDeviceForm
onSubmit={(values) => createDevice({variables: {values}})}
/>
)}
</Mutation>
)}
</ApolloConsumer>
)
In this example, a view shows a list of Organization
s, each containing a
sublist of User
s. When one or more users is removed from an organization,
it makes the following call:
refetch(client, [
['User', userIds],
['Organization', organizationId],
])
Passing an array to refetch
means to only refetch queries containing all of
the conditions in the array. So the query below would be refetched, but a query
containing only Organizations
or a query containing only User
s would not.
import * as React from 'react'
import gql from 'graphql-tag'
import refetch from 'apollo-magic-refetch'
import {Mutation, ApolloConsumer} from 'react-apollo'
import OrganizationView from './OrganizationView'
const query = gql`
query {
Organizations {
id
name
Users {
id
username
}
}
}
`
const mutation = gql`
mutation removeUsersFromOrganization($organizationId: Int!, $userIds: [Int!]!) {
result: removeUsersFromOrganization(organizationId: $organizationId, userIds: $userIds) {
organizationId
userIds
}
}
`
const OrganizationViewContainer = ({organization: {id, name, Users}}) => (
<ApolloConsumer>
{client => (
<Mutation
mutation={mutation}
update={(cache, {data: {result: {organizationId, userIds}}}) =>
refetch(client, [
['User', userIds],
['Organization', organizationId],
])
}
>
{removeUsersFromOrganization => (
<OrganizationView
organization={organization}
onRemoveUsers={userIds => removeUsersFromOrganization({
variables: {organizationId, userIds},
})}
/>
)}
</Mutation>
)}
</ApolloConsumer>
)
const OrganizationsViewContainer = () => (
<Query query={query}>
{({data}) => {
const {Organizations} = data || {}
if (!Organizations) return <div />
return (
<div>
<h1>Organizations</h1>
{Organizations.map((organization) => (
<OrganizationViewContainer
key={organization.id}
organization={organization}
/>
)}
</div>
)
}}
</Query>
)
Assuming the same Organization
s/User
s schema as above, the example performs
the necessary refetches when a user is created and added to an organization:
refetch(client, [['User'], ['Organization', organizationId]])
In this case no ids
are given for User
, so any query containing the an
Organization
with the given organizationId
in its results and selecting any
User
s would be refetched. (This doesn't perfectly exclude cases that fetch
Users and Organizations separately, instead of one nested inside the other, but
it's better than nothing).
import * as React from 'react'
import gql from 'graphql-tag'
import refetch from 'apollo-magic-refetch'
import { Mutation, ApolloConsumer } from 'react-apollo'
import CreateUserForm from './CreateUserForm'
const mutation = gql`
mutation createUser($organizationId: Int!, $values: CreateUser!) {
result: createUser(organizationId: $organizationId, values: $values) {
organizationId
id
username
}
}
`
const CreateUserFormContainer = ({ organizationId }) => (
<ApolloConsumer>
{(client) => (
<Mutation
mutation={mutation}
update={() =>
refetch(client, [['User'], ['Organization', organizationId]])
}
>
{(createUser) => (
<CreateUserForm
onSubmit={(values) =>
createUser({
variables: { organizationId, values },
})
}
/>
)}
</Mutation>
)}
</ApolloConsumer>
)
import refetch from 'apollo-magic-refetch'
Scans active queries in the given ApolloClient
and refetches any that contain
data matching the given type(s)/id(s).
The ApolloClient
in which to scan active queries.
The __typename
of the GraphQL type that was created or deleted, or an array of
[typename, predicate, idField]
tuples (predicate
and idField
are optional). If an
array is given, a query must match all of the conditions in the array to be
refetched.
A single id, an array of ids, or a Set
of ids that were deleted, or a
predicate function that takes an instance of the GraphQL type and returns true
if the query should be refetched. If given, only active queries whose current
result matches the predicate or contains an object with the given typename
and
id
will be refetched.
The name of the id field in the type that was deleted. This is only used if
predicate
is not an id, array, or Set
of ids, rather than a function
.
Prefetches type metadata by running an introspection query on the given on
ApolloClient
. The server must support client introspection queries;
otherwise use refetch.setTypeMetadata
.
The client to fetch type metadata from.
Sets the type metadata to use for determing which queries to refetch. Use this method if your server forbids client introspection queries.
The result of executing the typesQuery
GraphQL query
or a Promise
that will resolve to the result.
import { typesQuery } from 'apollo-magic-refetch'
The parsed GraphQL introspection query that gets all of the type metadata
needed to determine which queries to refetch. Use this if your server forbids
client introspection queries; execute this query on the server side and send
the result to the client code that calls
refetch.setTypeMetadata
.